dev-util/Tensile: add 6.1.1

Changes:
* fix USE=test dependency for dev-python/joblib (in 5.7.1 and 6.1.1)
* ReplacementKernels-cov3 directory does not exist anymore
* update expand-isa-compatibility patch to not use removed gcnArch field
* update llvm eclass to r1
* add LLVM_COMPAT 18

Signed-off-by: Sv. Lockal <lockalsash@gmail.com>
Signed-off-by: Sam James <sam@gentoo.org>

5 files changed, 203 insertions, 0 deletions

diff --git a/dev-util/Tensile/Manifest b/dev-util/Tensile/Manifest
index 4d93ffa204a1..fb95e0c702e7 100644
--- a/dev-util/Tensile/Manifest
+++ b/dev-util/Tensile/Manifest
@@ -2,3 +2,4 @@ DIST Tensile-5.0.2-PR1419.patch.gz 43754 BLAKE2B 3b1d7333ed0115f77816806e7735e25
 DIST rocm-Tensile-5.1.3.tar.gz 14357517 BLAKE2B 7c5881cc2b1d25db39f54a99f9f0a08395a5f7a58deebe74d9868e8f7044db5ae2f3907196adf16dc6abeba0f1f47f5cce5c8900891276daff8e4ad614cf3b94 SHA512 4405cb6de65c1da1fd93910fc901402a24e996ff04eccadf483c25d8eb518e03ea9bbf00e372dd5fe377696e71c39355603095f3c5fc1e26d311ed15d8e3b734
 DIST rocm-Tensile-5.4.2.tar.gz 14413805 BLAKE2B ca1c13785164439f6fe83f16b36b7a0b06a63b70b9f0741d77b0fff98f38d1e1c7e9a54ef61897e6eb0a4eaf52ec9e5f14f42a7e6fd657f26306afa3de0c74a2 SHA512 d99736a202c7353979e732941199b4b20e53ff2b6c19a7175609440a840aa5624bc53a8013c435f31ac37933527771521ff02137c1c8fd53c5317eca8533be80
 DIST rocm-Tensile-5.7.1.tar.gz 12396002 BLAKE2B ec2c2bc3f07fe22abd800b642c998ea80b9e4e2d0d28c4763d574dbd2fcb14c735223ddc7571d4ab3a48f0bd819cc835c16df9882049c5bf56f25ef729e6d654 SHA512 8f356f36cfec74ee9b000d1180cf52077dc5c595e5c5a40792fc3eb13f3139980c7440f9bc912218f946cf8137e2481d23fe39f110511c1bb55fe85a933ffd36
+DIST rocm-Tensile-6.1.1.tar.gz 12675289 BLAKE2B dfe3698d7dda14d5b20a6fe1ef6535ac9737ac330b71678a3a94e939adc2350e2dcf33e58c711ee90dae543efba59fe210bfa39fc69b63910ec4537db3f209bf SHA512 cf8b014b1ff7cd5d45236b8a9a6f5604b02e34a9245a163c672847b535d565151bf7e3a5127601681e80e6fe26964a0e544594044efe2fe68e39b0f1128f96e2
diff --git a/dev-util/Tensile/Tensile-5.7.1-r2.ebuild b/dev-util/Tensile/Tensile-5.7.1-r2.ebuild
index 94b4aac75b1b..64902c86549c 100644
--- a/dev-util/Tensile/Tensile-5.7.1-r2.ebuild
+++ b/dev-util/Tensile/Tensile-5.7.1-r2.ebuild
@@ -38,6 +38,7 @@ BDEPEND="
 		dev-python/pytest-forked[${PYTHON_USEDEP}]
 		dev-python/pytest-xdist[${PYTHON_USEDEP}]
 		dev-python/filelock[${PYTHON_USEDEP}]
+		dev-python/joblib[${PYTHON_USEDEP}]
 	)
 "
 
diff --git a/dev-util/Tensile/Tensile-6.1.1.ebuild b/dev-util/Tensile/Tensile-6.1.1.ebuild
new file mode 100644
index 000000000000..fcb77f77972d
--- /dev/null
+++ b/dev-util/Tensile/Tensile-6.1.1.ebuild
@@ -0,0 +1,136 @@
+# Copyright 1999-2024 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+EAPI=8
+
+PYTHON_COMPAT=( python3_{10..12} )
+DISTUTILS_USE_PEP517=setuptools
+ROCM_VERSION=${PV}
+LLVM_COMPAT=( 18 )
+
+inherit cmake distutils-r1 llvm-r1 prefix rocm
+
+DESCRIPTION="Stretching GPU performance for GEMMs and tensor contractions"
+HOMEPAGE="https://github.com/ROCmSoftwarePlatform/Tensile"
+SRC_URI="https://github.com/ROCmSoftwarePlatform/Tensile/archive/rocm-${PV}.tar.gz -> rocm-Tensile-${PV}.tar.gz"
+S="${WORKDIR}/${PN}-rocm-${PV}"
+
+LICENSE="MIT"
+SLOT="0/$(ver_cut 1-2)"
+KEYWORDS="~amd64"
+IUSE="client test"
+REQUIRED_USE="client? ( ${ROCM_REQUIRED_USE} )"
+
+RESTRICT="!test? ( test )"
+
+RDEPEND="${PYTHON_DEPS}
+	client? ( dev-libs/boost )
+	>=dev-cpp/msgpack-cxx-6.0.0
+	dev-python/pyyaml[${PYTHON_USEDEP}]
+	dev-python/msgpack[${PYTHON_USEDEP}]
+	=dev-util/hip-6*
+	>=dev-util/rocm-smi-4.3.0
+	$(llvm_gen_dep '
+		sys-devel/clang:${LLVM_SLOT}
+	')
+"
+DEPEND="${RDEPEND}"
+BDEPEND="
+	test? (
+		dev-python/pytest-forked[${PYTHON_USEDEP}]
+		dev-python/pytest-xdist[${PYTHON_USEDEP}]
+		dev-python/filelock[${PYTHON_USEDEP}]
+		dev-python/joblib[${PYTHON_USEDEP}]
+	)
+"
+
+distutils_enable_tests pytest
+
+PATCHES=(
+	"${FILESDIR}"/${PN}-4.3.0-output-commands.patch
+	"${FILESDIR}"/${PN}-5.4.2-fix-arch-parse.patch
+	"${FILESDIR}"/${PN}-5.4.2-use-ninja.patch
+	"${FILESDIR}"/${PN}-6.1.1-fix-msgpack-dependency.patch
+	"${FILESDIR}"/${PN}-6.0.2-expand-isa-compatibility.patch
+)
+
+CMAKE_USE_DIR="${S}/${PN}/Source"
+
+src_prepare() {
+	distutils-r1_src_prepare
+	sed -e "s,\@LLVM_PATH\@,$(get_llvm_prefix),g" \
+		"${FILESDIR}"/${PN}-5.7.1-gentoopath.patch > "${S}"/gentoopath.patch || die
+	eapply $(prefixify_ro "${S}"/gentoopath.patch)
+
+	pushd ${PN} || die
+
+	sed -e "/ROCM_SMI_ROOT/s,lib,$(get_libdir)," \
+		-i Source/cmake/FindROCmSMI.cmake || die
+	sed -r -e "/TENSILE_USE_LLVM/s/ON/OFF/" \
+		-i Source/CMakeLists.txt || die
+
+	# ${Tensile_ROOT}/bin does not exists; call command directly
+	sed -e "s,\${Tensile_ROOT}/bin/,,g" -i cmake/TensileConfig.cmake || die
+
+	local Tensile_share_dir="\"${EPREFIX}/usr/share/${PN}\""
+	sed -e "/HipClangVersion/s/0.0.0/$(hipconfig -v)/" -i Common.py || die
+
+	sed -e "s,os.path.dirname(os.path.realpath(__file__)),${Tensile_share_dir},g" \
+		-i ReplacementKernels.py Common.py ${PN}.py || die
+
+	sed -e "s|os\.path\.dirname.*$|\"${EPREFIX}/usr/share/Tensile/Source\", end='')|" -i __init__.py || die
+
+	popd || die
+
+	sed -e "/package_data/d" -e "/data_files/d" -i setup.py || die
+	use client && PATCHES= cmake_src_prepare  # do not apply patches again in cmake_src_prepare
+}
+
+src_configure() {
+	distutils-r1_src_configure
+	if use client; then
+		local mycmakeargs=(
+			-DCMAKE_SKIP_RPATH=ON
+			-DTENSILE_USE_MSGPACK=ON
+			-DTENSILE_USE_LLVM=ON
+			-DTensile_LIBRARY_FORMAT=msgpack
+			-DAMDGPU_TARGETS="$(get_amdgpu_flags)"
+		)
+		CXX=hipcc cmake_src_configure
+	fi
+}
+
+src_compile() {
+	distutils-r1_src_compile
+	use client && cmake_src_compile
+}
+
+python_install() {
+	distutils-r1_python_install
+
+	python_moduleinto Tensile
+	pushd Tensile || die
+	python_domodule Components
+	python_newexe Utilities/merge.py ${PN}-merge
+}
+
+src_install() {
+	distutils-r1_src_install
+
+	pushd ${PN} || die
+	insinto /usr/share/${PN}
+	doins -r Configs Perf Source CustomKernels
+	insinto /usr/$(get_libdir)/cmake/${PN}
+	doins cmake/*.cmake
+
+	if use client; then
+		pushd "${BUILD_DIR}" || die
+		dobin client/tensile_client
+	fi
+}
+
+# Test suite fails to start without this
+python_test() {
+	export ROCM_PATH="${EPREFIX}/usr"
+	epytest
+}
diff --git a/dev-util/Tensile/files/Tensile-6.0.2-expand-isa-compatibility.patch b/dev-util/Tensile/files/Tensile-6.0.2-expand-isa-compatibility.patch
new file mode 100644
index 000000000000..7a72d833bba0
--- /dev/null
+++ b/dev-util/Tensile/files/Tensile-6.0.2-expand-isa-compatibility.patch
@@ -0,0 +1,34 @@
+diff --git a/Tensile/Source/lib/source/hip/HipHardware.cpp b/Tensile/Source/lib/source/hip/HipHardware.cpp
+index ffa4f0f..f894366 100644
+--- a/Tensile/Source/lib/source/hip/HipHardware.cpp
++++ b/Tensile/Source/lib/source/hip/HipHardware.cpp
+@@ -24,6 +24,8 @@
+  *
+  *******************************************************************************/
+ 
++#include <cstring>
++
+ #include <Tensile/AMDGPU.hpp>
+ #include <Tensile/hip/HipHardware.hpp>
+ #include <Tensile/hip/HipUtils.hpp>
+@@ -55,6 +57,20 @@ namespace Tensile
+         {
+             hipDeviceProp_t prop;
+             HIP_CHECK_EXC(hipGetDeviceProperties(&prop, deviceId));
++            // coerce to compatible arch
++            int gcnArch = strncmp(prop.gcnArchName, "gfx", 3) == 0 ? std::stoi(prop.gcnArchName + 3) : 0;
++            switch(gcnArch)
++            {
++            case 902: case 909: case 912:
++                std::strcpy(prop.gcnArchName, "gfx900");
++                break;
++            case 1013:
++                std::strcpy(prop.gcnArchName, "gfx1010");
++                break;
++            case 1031: case 1032: case 1033: case 1034: case 1035: case 1036:
++                std::strcpy(prop.gcnArchName, "gfx1030");
++                break;
++            }
+ #if HIP_VERSION >= 50220730
+             int hip_version;
+             HIP_CHECK_EXC(hipRuntimeGetVersion(&hip_version));
diff --git a/dev-util/Tensile/files/Tensile-6.1.1-fix-msgpack-dependency.patch b/dev-util/Tensile/files/Tensile-6.1.1-fix-msgpack-dependency.patch
new file mode 100644
index 000000000000..7a3b1884c074
--- /dev/null
+++ b/dev-util/Tensile/files/Tensile-6.1.1-fix-msgpack-dependency.patch
@@ -0,0 +1,31 @@
+Gentoo uses "msgpack-cxx" as library name and adds important
+MSGPACK_NO_BOOST compile definition when msgpack-cxx us installed with USE=-boost
+--- a/Tensile/Source/lib/CMakeLists.txt
++++ b/Tensile/Source/lib/CMakeLists.txt
+@@ -98,24 +98,9 @@ if(TENSILE_USE_LLVM OR TENSILE_USE_MSGPACK)
+ endif()
+ 
+ if(TENSILE_USE_MSGPACK)
+-    find_package(msgpack REQUIRED NAMES msgpack msgpack-cxx msgpack-c)
++    find_package(msgpack-cxx REQUIRED)
+     target_compile_definitions(TensileHost PUBLIC -DTENSILE_MSGPACK=1)
+-
+-    if(TARGET msgpackc-cxx)
+-        get_target_property(msgpack_inc msgpackc-cxx INTERFACE_INCLUDE_DIRECTORIES)
+-    elseif(TARGET msgpackc)
+-        get_target_property(msgpack_inc msgpackc INTERFACE_INCLUDE_DIRECTORIES)
+-    elseif(TARGET msgpack-cxx)
+-        get_target_property(msgpack_inc msgpack-cxx INTERFACE_INCLUDE_DIRECTORIES)
+-    endif()
+-
+-    if(DEFINED msgpack_inc)
+-        # include C++ headers manually
+-        # External header includes included as system files
+-        target_include_directories(TensileHost 
+-            SYSTEM PRIVATE $<BUILD_INTERFACE:${msgpack_inc}>
+-        )
+-    endif()
++    target_link_libraries(TensileHost PRIVATE msgpack-cxx)
+ endif()
+ 
+ if(TENSILE_USE_LLVM)